ResNet101

Description

ResNet101 is a convolutional neural network architecture that belongs to the ResNet (Residual Network) family.With a total of 101 layers, ResNet101 comprises multiple residual blocks, each containing convolutional layers with batch normalization and rectified linear unit (ReLU) activations. These residual blocks allow the network to effectively capture complex features at different levels of abstraction, leading to superior performance on image recognition tasks.

Setup

Install

pip3 install onnx
pip3 install tqdm

Download

Pretrained model: https://download.pytorch.org/models/resnet101-63fe2227.pth

Dataset: https://www.image-net.org/download.php to download the validation dataset.

Model Conversion

python3 export.py --weight resnet101-63fe2227.pth --output resnet101.onnx

Inference

export DATASETS_DIR=/Path/to/imagenet_val/

FP16

# Accuracy
bash scripts/infer_resnet101_fp16_accuracy.sh
# Performance
bash scripts/infer_resnet101_fp16_performance.sh

INT8

# Accuracy
bash scripts/infer_resnet101_int8_accuracy.sh
# Performance
bash scripts/infer_resnet101_int8_performance.sh

Results